Ethereum vs Solana uptime, compared
The chain that has never halted, next to one that has, around seven times.
Incident history
The complete history of Ethereum incidents.
Ethereum mainnet has never halted. Plenty has gone wrong anyway. This is the long version of the timeline on the front page: every serious moment in more than a decade of block production, what actually happened, and why the chain kept going through each one.
What counts as “down”
Almost every argument about whether Ethereum has gone down is really an argument about definitions, so fix one before the list starts.
Down, on this site, means the network stopped producing blocks. Consensus stalled, the head stopped advancing, and a user could do nothing but wait for a coordinated restart. That is the definition that matters for a chain whose entire job is to keep ordering transactions, and by it Ethereum has been down for zero minutes since .
Everything below either left block production untouched or degraded the network without stopping it. One entry, the May 2023 finality delays, is the single case where a weaker guarantee than “producing blocks” lapsed for a short window. It is here precisely because it is the closest thing to a real halt in the record, and it still was not one.
The record, at a glance
Every serious moment in more than a decade, in order. Green means the chain produced blocks straight through; amber means it was degraded but never stopped.
- Jul 2015
Genesis
The first block is produced. The counter starts and has not reset since.
Baseline - Jun 2016
The DAO
A reentrancy bug drained ~$150M; the community hard-forked to reverse it. Blocks never stopped.
Chain live - Sep–Nov 2016
The spam war
Under-priced opcodes were abused until nodes crawled. Two live hard forks fixed it.
Degraded - Nov 2017
Parity multisig freeze
~500,000 ETH locked in a broken contract. The chain it lived on did not notice.
Contract, not chain - Mar 2020
Black Thursday
The worst congestion in the record; zero-bid liquidations. Every block still landed on schedule.
Congested - Nov 2020
Geth split & Infura outage
Un-upgraded nodes forked and much of the infra went dark. The canonical chain kept producing blocks.
Infra down - Sep 2022
The Merge
Proof of work to proof of stake, live, on a network securing hundreds of billions. Zero seconds of downtime.
Zero downtime - May 2023
The finality delays
Finality paused twice, ~25 and ~64 minutes, while blocks kept coming. The closest call, and still not a halt.
Finality paused - Mar 2024
Dencun
Blobs ship; layer-2 data costs fall by roughly an order of magnitude. A routine live upgrade.
No downtime - May 2025
Pectra
A batch of account and validator changes. Most users met it as a changelog, not an event.
No downtime - Today
Still going
An unbroken run of block production, roughly every twelve seconds, and counting.
Live now
2015 — Genesis, and the baseline
Ethereum’s first block was produced on at 15:26:13 UTC. No service-level agreement, no status page, no one to call. A handful of people ran client software on ordinary machines. From that block onward the chain has produced a new one roughly every twelve seconds, through everything that follows.
The baseline sets the bar for every incident after it. None of them reset the counter.
June 2016 — The DAO
On an attacker began draining The DAO, a large investment contract holding a substantial fraction of all circulating ETH, exploiting a reentrancy bug to siphon off roughly a third of a fund then worth around $150 million.
The response was the most divisive event in Ethereum’s history. The community chose to hard-fork the chain to reverse the theft, and not everyone agreed; the minority that rejected the reversal kept mining the original chain, which survives today as Ethereum Classic. What the crisis was not is a halt. Blocks never stopped. The disagreement got resolved at the social and governance layer, by people choosing which rules to run, not by halting the machine. The network produced every block throughout.
September–November 2016 — The spam war
Through the autumn of 2016, attackers hammered the network with transactions that exploited operations priced too cheaply relative to the work they demanded. Nodes slowed to a crawl, some ran out of memory and crashed, and block gas limits were dropped hard to cope.
The fix shipped in days, not months, and on the live chain. Two hard forks, Tangerine Whistle in October and Spurious Dragon in November, repriced the abused operations and cleared out the millions of empty accounts the attacks had created. This was a genuinely degraded period, the worst sustained attack on client performance the network had seen, and blocks never stopped landing.
November 2017 — The Parity multisig freeze
On , a user interacting with a shared library contract behind many Parity multisig wallets triggered a bug and then deleted the library, permanently locking roughly 500,000 ETH in the wallets that depended on it.
It was painful and expensive, and it belongs here for one reason: it is a textbook example of a failure people call “Ethereum” that had nothing to do with the network. A single smart contract broke. The chain the contract lived on did not notice and did not care. It kept producing blocks at the usual rate. The distinction between the network and the things built on top of it runs through this entire list.
March 2020 — Black Thursday
On , as markets crashed worldwide, ETH’s price halved in a day. Everyone tried to move at once. Gas prices spiked past 1,000 gwei, transactions that needed to land urgently could not, and in the most-cited casualty, MakerDAO collateral auctions cleared with zero-bid liquidations because keepers could not get transactions through the congestion.
This is the worst congestion event in Ethereum’s history, and exactly the kind of thing people mean when they say the network “fell over.” It did not. Every block was produced on schedule. The chain was full and expensive, which is a very different thing from being stopped.
November 2020 — The Geth chain split and the Infura outage
On , a transaction triggered a consensus bug that an earlier go-ethereum release had quietly patched. Nodes that had not upgraded followed a different chain from nodes that had. So much of the ecosystem’s infrastructure ran on the affected software that the fallout was large. Infura went down, and with it much of the wallet ecosystem: MetaMask and a number of exchanges briefly lost their view of the chain.
This is the incident most often misremembered as “Ethereum went down.” What went down was a slice of infrastructure. The canonical chain kept producing blocks the entire time; operators upgraded, and the network converged. If your wallet showed an error that day, the thing that failed was your route to Ethereum, not Ethereum. It is the same pattern the front page calls out in what people mean when they say Ethereum is down.
September 2022 — The Merge
On , Ethereum switched its consensus mechanism from proof of work to proof of stake, on the live network, while it secured hundreds of billions of dollars, with no maintenance window. Total downtime for the transition: zero seconds.
It is hard to overstate how unusual this is. The Merge is the engineering equivalent of changing an aircraft’s engines mid-flight, and for the network’s availability it was a non-event. A separate article covers why Ethereum upgrades cause no downtime; the Merge is the extreme case that proves the rule.
May 2023 — The finality delays
On and , the beacon chain twice stopped finalising, for roughly 25 minutes and then roughly 64 minutes, under an unusual load pattern that pushed some consensus clients past what they could keep up with in time.
This is the one entry that comes closest to a real outage, so the details matter. Blocks kept being produced. Transactions kept being included and processed. What lapsed was finality: the strongest settlement guarantee, the point past which a block is economically irreversible. For that window the chain was live but its firmest promise was running late, and it caught back up on its own without any coordinated restart. The closest call in the record, and still not a halt. The full explainer is here.
March 2024 — Dencun
On , the Dencun upgrade introduced blob-carrying transactions, cutting the cost of posting layer-2 data by roughly an order of magnitude overnight. Rollup fees fell dramatically. For the availability of mainnet it was, once again, a routine live upgrade with no downtime.
May 2025 — Pectra
On , the Pectra upgrade shipped a batch of changes to accounts and validators together. A hard fork on Ethereum is routine enough by now that most users learned about it from a changelog rather than from anything they noticed. No maintenance window, no downtime.
Today — still going
More than a decade in, the counter on the front page is not a promise about the future. It is a record of the past: an unbroken run of block production that has survived a $150M contract hack, a sustained denial-of-service campaign, the worst congestion the chain has ever seen, a client bug that split the network, and a live change of consensus mechanism.
The pattern
Read the list end to end and the incidents sort cleanly into three kinds, none of them a network halt:
- Application and contract failures. The DAO, the Parity freeze. Something built on Ethereum broke. The chain did not.
- Degradation under load. The 2016 spam war, Black Thursday. The network was slow, full or expensive, and kept producing every block.
- Infrastructure failures. The 2020 Infura outage. The route between users and the chain broke, while the chain ran underneath it.
The one genuine consensus-layer scare, May 2023, degraded a guarantee without stopping block production, then resolved itself. That is the record.
What actually does go down
None of this means nothing around Ethereum ever fails. Layer-2 rollups halt, RPC providers fail, wallets and exchanges have outages, and the ETH price moves for reasons that have nothing to do with the network’s availability. The point of keeping a strict definition of “down” is to be able to tell those apart. For how Ethereum’s record compares to a chain that has halted, see Ethereum vs Solana uptime.
Sources & further reading
- “The history of Ethereum,” ethereum.org, for an overview of the DAO fork, the 2016 forks, and the Merge.
- Ethereum Foundation Blog (blog.ethereum.org), for the original announcements of Tangerine Whistle, Spurious Dragon, The Merge, Dencun and Pectra.
- Go Ethereum release notes and the November 2020 chain-split post-mortems from the client and infrastructure teams involved.
- On-chain data via Etherscan and beaconcha.in for block and finality timing.
- Contemporaneous reporting on Black Thursday and the MakerDAO liquidations, and on the Parity multisig freeze.
Durations and figures are as publicly reported at the time of writing and are worth re-checking against primary sources before citing.
Is Ethereum producing blocks right now? The front page checks it live against several independent public nodes, every twelve seconds.